Studies in Inductive Logic and Probability Volume 1 Rudolf Carnap Editor Online Ebook Texxtbook Full Chapter PDF

Studies in Inductive Logic and
Probability Volume 1 Rudolf Carnap

(Editor)
Visit to download the full and correct content document:
https://ebookmeta.com/product/studies-in-inductive-logic-and-probability-volume-1-ru
dolf-carnap-editor/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
Studies in Inductive Logic and Probability Volume 2

Richard C. Jeffrey (Editor)
https://ebookmeta.com/product/studies-in-inductive-logic-and-
probability-volume-2-richard-c-jeffrey-editor/
40 Paradoxes in Logic Probability and Game Theory Presh

Talwalkar
https://ebookmeta.com/product/40-paradoxes-in-logic-probability-
and-game-theory-presh-talwalkar/
Studies in Medieval Philosophy Science and Logic Ernest

A. Moody
https://ebookmeta.com/product/studies-in-medieval-philosophy-
science-and-logic-ernest-a-moody/
American Studies in Europe, Volume 1: Their History and

Present Organization Sigmund Skard
https://ebookmeta.com/product/american-studies-in-europe-
volume-1-their-history-and-present-organization-sigmund-skard/
Bracha L Ettinger Matrixial Subjectivity Aesthetics
Ethics Volume 1 1990 2000 Volume 1 1990 2000 Studies in
the Psychosocial Ettinger
https://ebookmeta.com/product/bracha-l-ettinger-matrixial-
subjectivity-aesthetics-ethics-
volume-1-1990-2000-volume-1-1990-2000-studies-in-the-
psychosocial-ettinger/
California Studies in Classical Antiquity Volume 8

https://ebookmeta.com/product/california-studies-in-classical-
antiquity-volume-8-california-studies-in-classical-antiquity-
volume-8/

https://ebookmeta.com/product/california-studies-in-classical-
antiquity-volume-7-california-studies-in-classical-antiquity-
volume-7/
Christianity and the Eastern Slavs Volume 1 California

Slavic Studies Volume VIII Nicholas V. Riasanovsky
(Editor)
https://ebookmeta.com/product/christianity-and-the-eastern-slavs-
volume-1-california-slavic-studies-volume-viii-nicholas-v-
riasanovsky-editor/
Selected Collected Works Volume 2 Probability Theory

and Extreme Value Theory
https://ebookmeta.com/product/selected-collected-works-
volume-2-probability-theory-and-extreme-value-theory/
Studies in Inductive
Logic and Probability
VOLUME I
Studies in Inductive
Logic and Probability
VOLUME I
Rudolf Carnap and Richard C. Jeffrey
EDITORS
U N I V E R S I T Y OF C A L I F O R N I A PRESS
B E R K E L E Y • LOS A N G E L E S • L O N D O N • 1971
University of California Press
Berkeley and Los Angeles
University of California Press, Ltd.

London, England
Copyright © 1971 by The Regents of the University of California
ISBN: 0-520-01866-4
Library of Congress Catalog Card Number: 77-136025

Printed in the United States of America
Designed by Dave Comstock
Contents
Introduction 1
1. Inductive Logic and Rational Decisions by Rudolf Carnap 5
1. Decision Making 7
2. Actual Decisions 11
3. Rational Decisions 13
4. Credibility 17
5. Permanent Dispositions 21
6. Inductive Logic 25
7. The Question of Acceptance 29
2. A Basic System of Inductive Logic, Part I by Rudolf Carnap 33
1. Basic Concepts and Basic Axioms 35

A. Basic Concepts 35
B. The Basic Axioms 38
2. Individuals and Attributes 43
A. Monadic Predicate Languages 43
B. Stronger Languages 48
3. Models and Propositions 53
A. The Space of Models 53
B. Propositions 56
C. Sample Propositions and Random Variables 62
4. Pure and Applied Inductive Logic 69
A. The Distinction between Pure and Applied Logic 69
B. Requirements for Primitive Attributes 70
5. Basic Assumptions about Individuals, Attributes and Relations
77
6. Sublanguages 87
7. Regular 'g'-Functions 101
8. Coherent 'g'-Functions 105
9. Symmetric ^-Functions 117
10. One Family of Attributes 121
vi CONTENTS
11. Representative Functions for J t 131

A. MI-Functions 131
B. Ml-Sequences 140
C. MS- and MD-Functions 142
12. Representative Functions for r<i 151
13. The Principle of Instantial Relevance 161
3. Probability Measures and Integrals by Richard C. Jeffrey 167

1. Introduction 169
2. Measures 173
3. Measurable Functions 183
4. Integrals 187
5. Properties of the Integral 191
6. Lebesgue-Stieltjes Integrals 195
7. Extensions and Mixtures of Measures; Stationary, Symmetric,
and Bernoullian Measures 199
8. Indefinite Integrals and Derivatives 205
9. Conditioning: Probabilities and Expectations as Random
Variables 211
10. De Finetti's Representation Theorem 217
Bibliographical Remarks 220
Some Recent History 223
4. The Principle of Instantial Relevance by Jürgen Humburg 225
5. Applications of De Finetti's Theorem to Inductive Logic by Haim

Gaifman 235
Postscript Concerning Extension of Probability Functions 246
Selected Bibliography 253

Introduction
Carnap's Logical Foundations of Probability [1950] was planned as

the first in a two-volume work, Probability and Induction. A summary
of the system of inductive logic projected for Volume II, based on the
function c*, appeared as an appendix to Volume I; but two years later,
with the publication of The Continuum of Inductive Methods, it became
apparent that Volume II would not simply be the theory of c*. The
method based on that function appeared as a single point in the
continuum, distinguished by a combination of plausibility and
simplicity, but no longer the clearly chosen inductive method. During
the following two years' work at the Institute for Advanced Study,
Carnap's ideas changed still further, partly as a result of his work with
John Kemeny: the continuum of inductive methods itself was seen as
too narrow, e.g., because none of the c-functions in the continuum
are adequately sensitive to analogy by similarity. These investigations
resulted in a new axiom system for c-functions, which appeared as
Appendix B in Carnap [1959]. (See the bibliography at the end of this
volume.) But in the light of further work, this system, too, seemed too
narrow: see Carnap [1964a] §§25 and 26.
Thus, in the decade following publication of Carnap [1950], an
accumulation of small steps resulted in a change of perspective on past
and future work. Volume II of Probability and Induction would be no
mere elaboration of the outline at the end of Volume I, but the outcome
of at least another decade's work, the shape of which could not yet be
clearly seen. Furthermore, the technical apparatus elaborated in Volume
I no longer seemed satisfactory, partly because of Carnap's extensive
use of mathematical tools like de Finetti's representation theorem,
which had not figured in his earlier work, and partly because of a
desire to formulate inductive logic in terms that had come to be
standard in mathematical probability theory and theoretical statistics,
where probabilities are attributed to "events" (or "propositions")
which are construed as sets of entities which can handily be taken to be
models, in the sense in which that term is used in logic.
Then, in 1960, Carnap drew up a plan of articles for Studies in
Inductive Logic and Probability—a surrogate for Volume II of the
2 INTRODUCTION
work projected in 1950, in the form of a nonperiodical journal in which

various authors would report work in progress toward an adequate
system of inductive logic. To begin, some twenty articles were pro-
jected, by Carnap, Kemeny, Jeffrey, and Gaifman. The first few were
planned to be completed in the course of a year or two; but they are
finally appearing only now. A version of Art. 1 has already been
published (Carnap [1962a]), and an earlier version of the Basic System
(Art. 2 here, with its continuation to follow in Volume 2) was distributed
in dittographed form in 1959-61 under the title, "An Axiom System of
Inductive Logic". Meanwhile, a number of important papers (by
Gaifman, Krauss, and others) that bear on the project have appeared
in various journals. Thus this volume appears too late to avoid
dispersion and the generation of dittographed arcana, and to that
extent it fails in its original purpose. But part of that original project
was apparently unrealistic. The more important part is realized (in this
volume and the next) in Carnap's detailed presentation of his basic
system of inductive logic as it now exists. It is basic first in the sense
that many issues requiring more thorough exploration are touched on
here only lightly; notes for papers on these issues exist in Carnap's
shorthand, and it is to he hoped that this material will begin to appear
in print in later volumes. And it is basic, too, in the sense that what we
have now is to be viewed as a foundation, not as a completed structure.
Nor, of course, is the foundation immune from revision: difficulties
may well appear in the upper stories which will force changes in the
basics.
Rudolf Carnap and Richard C. Jeffrey
1969
Postscript
Carnap died in September, 1970, after a brief illness. He had
finished revising all but the last two sections of the continuation of the
Basic System (Art. 2) which will appear in Volume II. These last two
sections are now being revised by Lary Kuhns and Gordon Matthews,
who had been working with Carnap over the years, and with whom
Carnap had discussed the two sections in detail.
The Introduction above is a draft, written by me and incompletely
revised by Carnap, who approved it and sent me some additional notes,
which are reproduced below essentially as he wrote them. They give
some of the further facts he wanted mentioned.
Thanks are due to Carnap's assistant, Brooks Colburn, for
proofreading Articles 1 and 2, and for other help; and to Hannaliese
INTRODUCTION 3
Thost-Carnap, for keeping her father's working papers together and

accessible to his collaborators.
Richard C. Jeffrey
February, 1971
Further Material for the Introduction
1952-1954: Carnap was at the Institute for Advanced Study, in

Princeton.
First year: 1952-1953. Throughout this period Carnap collaborated
regularly with John Kemeny. (Kemeny was teaching at the Department
of Philosophy at Princeton University. He was also a very efficient
mathematician; as such he had been an assistant to Einstein.) Kemeny
and Carnap worked chiefly on the problem of several families of
attributes. Carnap had found a solution for two families (see [1959]
Anhang B, sec. VIII). They now extended this to a language with any
(finite) number of families. Kemeny also studied de Finetti's work,
especially his representation theorem, according to which any
symmetric M-function can be represented by a distribution function in
the structure space (see Art. 3, §10). Sometimes they also had talks
with L. J. Savage. Among other things, Savage showed them that the use
of a language L v with a finite number N of individuals is not advis-
able, because a symmetric M-function in LN cannot always be extended
to an M-function in a language with a greater number of individuals.
Yehoshua Bar-Hillel was in Princeton for some time in April, 1953.
Carnap told him about his ideas on a semantical concept of content
measure or amount of information based on the logical concept of
probability, in contrast to the statistical concept of amount of informa-
tion (Shannon). This led to the Research Report: Carnap and Bar-
Hillel [1952]. Frequently there is confusion between these two concepts.
Only the second, not the first, is related to the physical concept of
entropy. Bar-Hillel and Carnap also had some talks with John von
Neumann. He asserted that the basic concepts of quantum theory are
subjective and that this holds especially for entropy since this concept is
based on probability and amount of information. They tried in vain to
convince him of the existence of a difference in each of these two pairs
of concepts (see Bar-Hillel [1964, p. 12]).
Second Year: 1953-1954. Kemeny was in England. Carnap
worked mostly on the big manuscript on entropy (unpublished). He
tried to clarify the nature of the entropy concept as it occurs in statistical
mechanics (Boltzmann and Gibbs); and especially to show the difference
between this physical concept and a corresponding logical concept.
4 INTRODUCTION
Carnap had many talks with Hilary Putnam; also some talks with Kurt
Godel. [I hope to provide more material on the second year later, when
I shall have read more of my diary.]
1954-1962: Carnap was teaching at UCLA. He had many talks
with philosophers and others who came to Los Angeles for a time,
often for work at the Rand Corporation.
Kemeny and Richard Jeffrey were Research Associates in Carnap's
research grant NSF (2) for 1959-62. Jeffrey had written his doctoral
dissertation on inductive logic with Hempel.
Research Assistants
Gordon Matthews and Lary John Kuhns (financed by Research
Committee of UCLA 1957-63; later by NSF grants (3) and (4), see
below).
Haim Gaifman, 1959-60 (NSF grant (2)).
Peter Krauss, 1962-63 (NSF grant (3)).
Institutions and Foundations
Institute for Advanced Study, Princeton, 1952-54. Stipend from the
institute, supplemented in the first year by:
Bollingen Foundation, 1952-53, and NSF (1), and in the second
year by:
John Simon Guggenheim Memorial Foundation, 1953-54.
UCLA Research Committee, 1957-62: yearly grants for Research
Assistants Matthews and Kuhns.
Office of Naval Research: Contract Nonr-233(55). July 1, 1961-
January 31, 1963, only for Research Assistant Krauss.
National Science Foundation:
(1) Grant G163.
(2) Grant G7585.
(3) Grant G22315.
(4) Grant GS550; this is a continuation of (3).
I wish to express my deep appreciation for the support of my
research work by the Institute and the foundations listed above.
Rudolf Carnap
May 21, 1970
1
Inductive Logic and

Rational Decisions
BY R U D O L F C A R N A P
1
Decision Making1
By "inductive logic" I understand a theory of logical probability

providing rules for inductive thinking. I shall try here to make clear
the nature of inductive logic by showing how it can be used in deter-
mining rational decisions.
Decision theory involves the concepts of utility and probability.
I shall try to show that, in this context, we must understand "probability"
not as relative frequency, but as degree of belief. This is originally, in
descriptive decision theory, a psychological concept, referring to actual
beliefs of actual human beings. Later I shall go to normative decision
theory by introducing some requirements of rationality. Up to that
point I shall be in agreement with the representatives of a personal (or
"subjective") conception of probability. But then I shall take a further
step leading from a quasi-psychological to a logical concept of prob-
ability. By "inductive logic" I understand the theory of this concept.
Let us begin with the customary schema of decision making. A
person Xata certain time T has to make a choice among possible acts
Ax, A2,. ... We assume the following. A'knows that the possible states
of nature (or of a part of nature relevant for his decision) at time T are
W u JV 2 ,.. .; but he does not know which of them is the actual state.
The number of possible acts and the number of possible states are
finite. X knows the following: if he were to carry out the act Am and if
the state Wn were the actual state of nature, then the outcome would be
O m n . This outcome Om n is uniquely determined by Am and Wn; and X
knows how it is determined. We assume that there is a utility function
Ux for the person X and that X knows his utility function so that he
can use it in order to calculate values.
Now we define the (subjective) value (desirability) of a possible act
Am for Z at time T:
(1) Definition. Vx,T(Am) = £ Wx{Om,n)P{Wn)], where P{Wn)
n
is the probability of the state W n , and the sum covers all possible states
W .
rr n*
In other words, we take as the value of the act Am for X the
expected utility of the outcome of this act. (1) holds for the time T
before any of the possible acts is carried out. It refers to the contempla-
ted act Am; therefore it uses the utilities for the possible outcomes
1
This article is a modified and expanded version of my paper "The aim of
inductive logic" which appeared in: Logic, methodology and philosophy of science:
Proceedings of the 1960 International Congress, ed. E. Nagel, P. Suppes, and A. Tarski
(Stanford University Press, 1962).
8 STUDIES IN INDUCTIVE LOGIC AND PROBABILITY
0 m , „ of act Am in the various possible states fV„. [If the situation is

such that the probability of Wn could possibly be influenced by the
assumption that act A m were carried out, we should take the condi-
tional probability P(Wn | Am) instead of P{Wn). Analogous remarks
hold for our later forms of the definition of V.]
Now the customary Bayesian rule of decision making says:
(2) Choose an act so as to maximize the value V.
As a principle (statement) it may be stated in either of the following
two forms, which are essentially different but are not always clearly
distinguished:
(3a) Decisions are (usually, under normal conditions) made in
such a way that the chosen act has the maximum value.
(3 b) A rational decision consists in the choice of an act that has
the maximum value.
The principle (3a) is an alleged psychological law belonging to
descriptive decision theory, which is a part of psychology. In contrast,
(3b) belongs to normative decision theory, which states conditions of
rationality for decisions. This distinction will presently be discussed.
But first we have to remove an ambiguity in the definition (1) of value,
concerning the interpretation of the probability P. There are several
conceptions of probability; thus the question arises which of them is
adequate in the context of decision theory.
The main conceptions of probability are often divided into two
kinds, statistical (or objective) and personal (or subjective) concep-
tions. As I see it, these are not two incompatible doctrines concerning
the same concept, but rather two theories concerning two different
probability concepts, both of them legitimate and useful. The concept
of statistical probability is closely connected with relative frequencies
in mass phenomena. It plays an important role in mathematical stat-
istics, and it occurs in laws of various branches of empirical science,
especially physics.
The second concept is personal probability. It is the probability
assigned to a proposition or event H by a person X, in other words, the
degree of belief of X in H. Now it seems to me that we should clearly
distinguish two versions of personal probability, one representing the
actual degree of belief and the other the rational degree of belief.
Which of these two concepts of probability, the statistical or the
personal, ought to be used in the definition of value and thereby in the
decision principle? At the present time, the great majority of those who
work in mathematical statistics still regard the statistical concept of
probability as the only legitimate one. This concept refers, however,
to an objective feature of nature, a feature that holds whether or not the
observer X knows about it. And in fact, the numerical values of
INDUCTIVE LOGIC AND RATIONAL DECISIONS 9
statistical probability are in general not known to X. Therefore this con-

cept is unsuitable for a decision principle. It seems that for this reason a
number of those who work in decision theory, be it descriptive or
normative, incline toward the view that some version of the personal
concept of probability must be used here. I agree with this view.
The statistical concept of probability remains, of course, a legiti-
mate and important concept both for mathematical statistics and for
many branches of empirical science. And in the special case that X
knows the statistical probabilities for the relevant states Wn but does
not know which is the actual state, the decision principle would use
these probabilities. There is general agreement on this point. And this
is not in conflict with the view that the decision principle should refer to
personal probability, because in this special situation the personal
probability for X would be equal to the statistical probability.
Once we recognize that decision theory needs the personal con-
cept of probability, it is clear that the theory of actual decisions involves
the first version of this concept, i.e., the actual degree of belief, and the
theory of rational decisions involves the second version, the rational
degree of belief.
2
Actual Decisions
Let us first discuss descriptive decision theory, the theory of

actual decisions. The concept of probability in the sense of the actual
degree of belief is a psychological concept; its laws are empirical laws
of psychology, to be established by the investigation of the behavior of
persons in situations of uncertainty, e.g., behavior with respect to bets
or games of chance. I shall use for this psychological concept the
technical term degree of credence or credence for short. In symbols,
I write 1CTx t ( H ) ' for "the [degree of] credence of the proposition H
for the person X at the time T." Different persons X and Y may have
different credence functions Crx T and C r Y T . And the same person X
may have different credence functions C r x § T i and Cr A -,r 2 at different
times T1 and T2; e.g., if X observes between 7\ and T2 that H holds, then
CTX TI(H) CTX T2(H). (Let the ultimate possible cases be repre-
sented by the points of a logical space, usually called the probability
space. Then a proposition or event is understood, not as a sentence, but
as the range or truth set of a sentence, i.e., the set of points representing
those possible cases in which the sentence holds. To the conjunction of
two sentences corresponds the intersection of the propositions.)
On the basis of credence, we can define conditional credence,
"the credence of H with respect to the proposition E" (or " . . . given
E"):
. CRX T(E N H)
(4) Definition. CXX T{H\ E) = ——^ —-—, provided that
CXX T{E) > 0. Cr'X,T{H | E) is the credence that H would have for X
at TIF X ascertained that E holds.
Using the concept of credence, we now replace the definition of
value (1) by the following:
(5) Definition. VX T(Am) = £ [Ux{Om,n)Cvx_T{Wn)].
n
As was pointed out by Frank P. Ramsey, we can determine X's
credence function by his betting behavior. A bet is a contract of the
following form: Xpays into the pool the amount u, his partner Y pays
the amount v; they agree that the total stake u + v goes to X if the
hypothesis H turns out to be true, and to Y if it turns out to be false.
If X accepts this contract, we say that he bets on H with the total stake
u + v and with the betting quotient q = uj(u + v) (or, at odds of u to v).
If we apply the decision principle with the definition (5) to the situation
in which X may either accept or reject an offered bet on H with the
betting quotient q, we find that X will accept the bet if q is not larger
than his credence for H. Thus we may interpret CrX T(H) as the
highest betting quotient at which X is willing to bet on H. (As is

well known, this holds only under certain conditions and only
approximately.)
Utility and credence are psychological concepts. The utility
function of X represents the system of valuations and preferences of X;
his credence function represents his system of beliefs (not only the
content of each belief, but also its strength). Both concepts are theore-
tical concepts which characterize the state of mind of a person; more
exactly, the nonobservable microstate of his central nervous system,
not his consciousness, let alone his overt behavior. But since his behav-
ior is influenced by his state, we can indirectly determine characteristics
of his state from his behavior. Thus experimental methods have been
developed for the determination of some values and some general
characteristics of the utility function and the credence function ("sub-
jective probability") of a person on the basis of his behavior with respect
to bets and similar situations. Interesting investigations of this kind have
been made among others by F. C. Mosteller and P. Nogee, and m o r e
recently by D. Davidson, P. Suppes, and S. Siegel [1957] (see the
Bibliography at the end of this volume).
3
Rational Decisions
Now we go from descriptive to normative decision theory. The

latter is of greater interest to us, not so much for its own sake (its
methodological status is in fact somewhat problematic), but because
it is the connecting link between descriptive decision theory and
inductive logic. Normative decision theory is concerned not with actual
credence, but with rational credence. (We should also distinguish here
between actual utility and rational utility; but we omit this.) The state-
ments of a theory of this kind are not found by experiments but are
established on the basis of requirements of rationality. The formal pro-
cedure usually consists in deducing theorems from axioms that are
justified by general considerations of rationality, as we shall see. It
seems fairly clear that the probability concepts used by the following
authors are meant in the sense of rational credence (or of rational
credibility, which I shall explain presently): John Maynard Keynes
[1921], Frank P. Ramsey [1931a], Harold Jeffreys [1957], B. O. Koop-
man [19406], Georg Henrik von Wright [1957], I. J. Good [1950],
Leonard J. Savage [1954],
It is now clear that the probability concept of Bruno de Finetti
([1937] (English translation [1964])) belongs to the same kind. His
earlier formulations were sometimes misleading, e.g., when he said
that his probability concept refers not to rational, but to actual beliefs.
He made his view clear in a lengthy footnote on p. I l l of Kyburg-
Smokler [1964], At the end of this note he says: "In order to avoid
frequent misunderstandings it is essential to point out that probability
theory is not an attempt to describe actual behavior; its subject is
coherent behavior, and the fact that people are only more or less
coherent is inessential."
[Some remarks on terminology. Instead of "statistical" and
"personal" as characterizing adjectives attached to "probability," the
terms "objective" and "subjective" are frequently used. Now the
concept of statistical probability can indeed be correctly characterized
as being objective. But the use of "subjective" for the concept of
personal probability seems to me highly questionable. We might still
apply it to the concept of actual degree of belief, but its use for the
rational degree of belief in normative decision theory would seem to me
inappropriate, and still more for the logical concept of probability to
be discussed later. (See my remarks in [1950], p. 238; and ibid., p. 43,
with a quotation from J. M. Keynes [1921], p. 4.) I agree with Keynes
in the view that the concept of logical probability is just as objective as
the concept of logical implication in deductive logic. (This logical

objectivity is, of course, quite different from the factual objectivity of
concepts in physics and of statistical probability.) The term "personal
probability" was introduced by L. J. Savage [1954]; for the reasons
mentioned above, it seems to me preferable to "subjective probability,"
especially in normative decision theory.]
I now give some examples of rationality requirements that I regard
as valid. There is no general agreement in the question of validity.
Each author regards as "rational credence functions" those that
satisfy the rationality requirements accepted by him. Sometimes, to
make the picture more lively, one speaks of a "completely rational"
person X; this means an (imaginary) person whose credence function
is assumed to be perfectly rational.
Suppose that X makes n simultaneous bets; let the /th bet (/' =
1,. . . , n) be on the proposition Ht with the betting quotient qt and the
total stake st. Before we observe which of the propositions Hi are true
and which are false, we can consider the possible cases. For any possible
case, i.e., a logically possible assignment of truth values to the n
propositions H{, we can calculate the gain or loss for each bet and hence
the total balance of gains and losses from the n bets. If in every possible
case Xsuffers a net loss, i.e., his total balance is negative, it is obviously
unreasonable for X to make these n bets. Let X's credence function at a
given time be Cr. By a (finite) betting system in accordance with Cr we
mean a finite system of n bets on n arbitrary propositions H{ (i =
1 , . . . , « ) with n arbitrary (positive) stakes s(, but with the betting
quotients qt = C r ( H { ) .
(6) Definition. A function Cr is coherent if and only if there is no
betting system in accordance with Cr such that there is a net loss in
every possible case.
For X to make bets of a system of this kind would obviously be
unreasonable. Therefore we lay down the first requirement as follows:
Rl. In order to be rational, Cr must be coherent.
Now the following important result holds:
(7) A function Cr from propositions to real numbers is coherent
if and only if Cr is a probability measure, i.e., satisfies the basic axioms
of the calculus of probability. (7) was first proved by de Finetti [1931].
Let Cr' be the conditional credence function defined on the basis
of Cr by (4). As ordinary bets are based on Cr, conditional bets are
based on Cr'. The concept of coherence can be generalized so as to be
applicable also to conditional credence functions. (7) can then easily
be extended by the result that a conditional credence function Cr' is
coherent if and only if Cr' satisfies the customary basic axioms of
conditional probability, including the general multiplication axiom.
Following Shimony [1955], we introduce now a concept of coher-

ence in a stronger sense, for which I use the term "strict coherence":
(8) Definition. A function Cr is strictly coherent if and only if Cr is
coherent and there is no (finite) system of bets in accordance with Cr
on molecular propositions such that the result is a net loss in at least
one possible case, but not a net gain in any possible case.
It is clear that it would be unreasonable to accept the bets of a
system of the kind just specified. Therefore we lay down the second
requirement:
R2. In order to be rational, a credence function must be strictly
coherent.
We define regular credence function (essentially in the sense of
Carnap [1950, §55,4]):
(9) Definition. A function Cr is regular if and only if Cr is a prob-
ability measure and, for any molecular proposition H, C r ( H ) = 0 only
if H is impossible.
By analogy with (7), we have now the following important theorem
(10); its first part (". . . regular i f . . .") is attributable to Shimony
[1955], its second part (". . . only i f . . . " ) was proved by John Kemeny
[1955] and by R. Sherman Lehman [1955] independently of each other.
(10) A function Cr is strictly coherent if and only if Cr is regular.
Most of the authors of systems for personal or logical probability
adopt only the basic axioms; they thus require nothing but coherence.
A few go one step further by including an axiom for what I call regu-
larity; they thus require in effect strict coherence, but nothing more.
Axiom systems of both kinds are extremely weak; they yield no result
of the form "P(H \ E) = r," except in the trivial cases where r is 0 or 1.
In my view, inductive logic should accomplish much more.
The two preceding requirements apply to any credence function
that holds for X at any time T of his life. We now consider two of these
functions, Crre for the time Tn and Cr n + 1 for a time Tn+1 shortly after Tn.
Let the proposition E represent the observation data obtained by X
between these two time points. The third requirement refers to the
transition from Cr„ to Cr„ + 1 :
R3 (a) The transformation of Cr„ into Cr n + 1 depends only on the
proposition E.
(b) More specifically, Cr„ + 1 is determined by Cr n and E as follows:
for any H, Ctn+1(H) = C r n ( E n tf)/Cr„(£) ( h ence = C^(H\E) by
definition (4)).
Part (a) is of course implied by (b). I have separated part (a) from
(b) because A"s function Cr might satisfy (a) without satisfying (b).
Part (a) requires merely that X be rational to the extent that changes in
his credence function are influenced only by his observational results,
but not by any other factors, e.g., feelings like his hopes or fears
concerning a possible future event H, feelings that in fact often influence
the beliefs of all actual human beings. Part (b) specifies exactly the
transformation of Cr„ into Cr n+1 ; the latter is the conditional credence
Cr'n with respect to E. The rule (b) can be used only if Cr n (E) ^ 0;
this condition is fulfilled for any possible observational result, provided
that Cr n satisfies the requirement of strict coherence.
4
Credibility
Let the proposition En+2 represent the data obtained between Tn+1
and a later time point Tn+2. Let Crn+2 be the credence function at Tn+2
obtained by R3(b) from Cr K + 1 with respect to En+2. It can easily be
shown that the same function Cr„ + 2 results if R3(b) is applied to Cr n
with respect to the combined data En+1 n En+2. In the same way we can
determine any later credence function C r „ + m from the given function
Cr„ either in m steps, applying the rule R3(b) in each step to one datum
of the sequence En+2,. . . , En+m, or in one step to the intersection
m
| J En+V. If m is large so that the intersection contains thousands of
3>=1
single data, the objection might be raised that it is unrealistic to think
of a procedure of this kind, because a man's memory is unable to
retain and reproduce at will so many items. Since, however, our goal is
not the psychology of actual human behavior in the field of inductive
reasoning, but rather inductive logic as a system of rules, we do not aim
at realism. We may make the further idealizing assumption that X is
not only perfectly rational but has also an infallible memory. Our
assumptions deviate from reality very much if the observer X is a
natural human being, but not so much if we think of X as a robot with
organs of perception, memory, data processing, decision making, and
acting. Thinking about the design of a robot might help us in finding
rules of rationality. Once found, these rules can be applied not only in
the construction of a robot but also in advising human beings in their
effort to make their decisions as rational as their limited abilities permit.
Consider now the whole sequence of data obtained by X up to the
present time Tn: Elt E2, • • • , En. Let J^x,T n or > f ° r short, Kn be the
proposition representing the combination of all these data:
(11) Definition. Kn = f | E{.

!= 1
Thus Kn represents, under the assumption of infallible memory,
the total observational knowledge of X at the time Tn. N o w consider
the sequence of A"s credence functions. In the case of a human being
we would hesitate to ascribe to him a credence function at a very early
time point, before his abilities of reason and deliberate action are
sufficiently developed. But again we disregard this difficulty by thinking
either of an idealized human baby or of a robot. We ascribe to him a
credence function Cr x for the time point 7 \ ; C r t represents X's personal
probabilities based upon the datum Ex as his only experience. Going
even one step further, let us ascribe to him an initial credence function
Cr 0 for the time point T0 before he obtains his first datum Ev Any later
function Cr„ for a time point Tn is uniquely determined by Cr 0 and Kn:
(12) For any H, C r n ( H ) = Cr' 0 (H \ Kn), where Cr; is the condi-
tional function based on Cr„.
C r n { H ) is thus seen to be the conditional initial credence of H
given Kn.
How can we understand the function Cr 0 ? In terms of the robot,
Cr 0 is the credence function that we originally build in and that he
transforms step for step, with regard to the incoming data, into the later
credence functions. In the case of a human being X, suppose that we
find at the time Tn his credence function Cr„. Then we can, under
suitable conditions, reconstruct a sequence Eu . . . , En, the proposition
Kn, and a function Cr 0 such that (a) Eu . . . ,En are possible observation
data, (b) Kn is defined by (11), (c) Cr 0 satisfies all requirements of
rationality for initial credence functions, and (d) the application of
(12) to the assumed function Cr 0 and Kn would lead to the ascertained
function Cr n . We do not assert that X actually experienced the
data Eu . . . , En, and that he actually had the initial credence function
Cr„, but merely that, under idealized conditions, his function Cr n
could have evolved from Cr 0 by the effect of the data El,.. . , En.
For the conditional initial credence function (Cr^) we shall also
use the term "credibility function" and the symbol 'Cred'. As an alter-
native to defining 'Cred' on the basis of 'Cr 0 ', we could introduce it as a
primitive term. In this case we may take the following universal state-
ment as the main postulate for the theoretical primitive term 'Cred':
(13) Let Cred be any function from pairs of propositions to real
numbers, satisfying all requirements which we have laid down or shall
lay down for credibility functions. Let H and A be any propositions
(A not empty). Let X be any observer and T any time point. If A"s
credibility function is Cred and his total observational knowledge at
Tis A, then his credence for H at 7" is Cred(//1 A).
Note that (13) is much more general than (12). There the function
Cred (or Cr^) was applied only to those pairs H, A, in which A is a
proposition of the sequence Kx, K2, . . . , and thus represents the actual
observational knowledge of A" at some time point. In (13), however, A
may be any nonempty proposition. Let A1 be a certain nonempty
proposition that does not occur in sequence Ku K2, . . . , and Hx some
proposition. Then the statement
C r T ( / f x ) = Cred(//j | A J
is to be understood as a counterfactual conditional as follows:
(14) Let X's credibility function be 'Cred'. If his total observa-
tional knowledge at some time point T had been Ar, then his credence
for Hx at T would have been equal to Cred(// X | At).
INDUCTIVE LOGIC A N D RATIONAL DECISIONS 19
This is a true counterfactual based on the postulate (13), analogous

to ordinary counterfactuals based on physical laws.
Applying (13) to A"s actual total observational knowledge Kx T
at time T, we have:
(15) For any H, CrXtT(H) = C r e d X ( H \ Kx/r).
Now we can use credibility instead of credence in the definition of
the value of an act A m , and thereby in the decision rule. Thus we have
instead of (5):
(16) Definition. Vx,T{Am) = £ [Ux{OmJCveAx{Wn |
n
[If the situation is such that the assumption of A m could possibly
change the credence of Wn, we have to replace 'Kx by LKX T O Am' \
see the remark in the paragraph following (1).]
If Cred is taken as primitive, Cr 0 can be defined as follows:
(17) Definition. For any H, C r 0 ( H ) = C r e d ( / / | Z), where Z is the
necessary proposition (the tautology).
This is the special case of (13) for the initial time T0, when A"s
knowledge K 0 is the tautology.
While CrX T characterizes the momentary state of X at time T
with respect to his beliefs, his function C r e d Y is a trait of his underlying
permanent intellectual character, namely his permanent disposition for
forming beliefs on the basis of his observations.
5
Permanent Dispositions
Since each of the two functions Cr 0 and Cred is definable on the

basis of the other, there are two alternative equivalent procedures for
specifying a basic belief-forming disposition, namely, either by Cr 0
or by Cred.
Most of those who have constructed systems of subjective prob-
ability (in the narrower sense, in contrast with logical probability),
e.g., Ramsey, de Finetti, and Savage, have concentrated their attention
on what we might call " a d u l t " credence functions, i.e., those of persons
sufficiently developed to communicate by language, to play games,
make bets, and so on, hence persons with an enormous amount of
experience. In descriptive decision theory it has many practical advan-
tages to take adult persons as subjects of investigation, since it is
relatively easy to determine their credence functions on the basis of
their behavior with games, bets, and the like. When I propose to take
as a basic concept, not adult credence but either initial credence or
credibility, I must admit that these concepts are less realistic and
remoter from overt behavior and may therefore appear as elusive and
dubious. On the other hand, when we are interested in normative
decision theory, these concepts have important methodological
advantages. Only for these concepts, not for credence, can we find a
sufficient number of requirements of rationality as a basis for the
construction of a system of inductive logic.
If we look at the development of theories and concepts in various
branches of science, we find frequently that it was possible to arrive at
powerful laws of widespread generality only when the development of
concepts, beginning with directly observable properties, had progressed
step by step to more abstract concepts, connected only indirectly with
observables. Thus, physics proceeds from concepts describing visible
motion of bodies to the concept of a momentary electric force, and then
to the still more abstract concept of a permanent electric field. In the
sphere of human action we have first concepts describing overt behavior,
say of a boy who is offered the choice of an apple or an ice cream cone
and takes the latter; then we introduce the concept of an underlying
momentary inclination, in this case the momentary preference of ice
cream over apple; and finally we form the abstract concept of an
underlying permanent disposition, in our example the general utility
function of the boy.
What I propose to do here is simply to take the same step—from
momentary inclination to the permanent disposition for forming
22 STUDIES IN INDUCTIVE LOGIC A N D PROBABILITY
momentary inclinations—also with the second concept occurring in the

decision principle, namely, personal probability or degree of belief.
Here it is the step from credence to credibility.
When we wish to judge the morality of a person, we do not simply
look at some of his acts; we study rather his character, the system of his
moral values, which is part of his utility function. Observations of
single acts without knowledge of motives give little basis for judgment.
Similarly, if we wish to judge the rationality of a person's beliefs, we
should not look simply at his present beliefs. Information on his
beliefs without knowledge of the evidence out of which they arose tells
us little. We must rather study the way in which the person forms his
beliefs on the basis of evidence. In other words, we should study his
credibility function, not simply his present credence function. F o r
example, let X have the evidence E that from an urn containing white
and black balls ten balls have been drawn, with replacement, two of
them white and eight black. Let Y have the evidence E' which is similar
to E, but with seven balls white and three black. Let H be the prediction
that the next ball drawn will be white. Suppose that for both X and Y
the credence of H is 2/3. Then we would judge this same credence
value 2/3 of the proposition H as unreasonable for X, but reasonable
for Y. We would condemn a credibility function Cred as nonrational if
C r e d ( H | E) = 2/3; while the result C r e d ( H \ E') = 2/3 would be no
ground for condemnation.
Suppose X has the credibility function Cred, which leads him, on
the basis of his knowledge Kn at time Tn, to the credence function Cr re ,
and thereby, with his utility function U, to the act Arn. If this act seems
to us unreasonable in view of his evidence Kn and his utilities, we shall
judge that Cred is nonrational. But for such a judgment on Cred it is
not necessary that X i s actually led to an unreasonable act. Suppose that
for E and H as in the above example, Kn contains E and otherwise
only evidence irrelevant f o r H . Then we have Crn(H) = C r e d ( H | Kn) =
C r e d ( H | E) = 2/3; and this result seems unreasonable on the given
evidence. If X bets on H with betting quotient 2/3, this bet is unreason-
able, even if he wins it. But his credence 2/3 is anyway unreasonable,
no matter whether he acts on it or not. It is unreasonable because there
are possible situations, no matter whether real or not, in which the
result C r e d ( H \ E) — 2/3 would lead him to an unreasonable act.
Furthermore, it is not necessary for our condemnation of the function
Cred that it actually leads to unreasonable Cr-values. Suppose that
another man X' has the same function Cred but is not led to the un-
reasonable Cr-value in the example, because he has an entirely different
life history, and at no time is his total knowledge either £ or a combina-
tion of E with data irrelevant for H. Then we would still condemn the
function Cred and the man X' characterized by this function. Our
argument would here, as frequently in an intellectual or moral con-
demnation, be expressed in a counterfactual conditional as follows:
if the total knowledge of X' had at some time been E, or E together
with irrelevant data, then his credence for H would have had the
unreasonable value 2/3. The same considerations hold, of course, for
the initial credence function Cr 0 corresponding to the function Cred;
for, on the basis of any possible knowledge proposition K, Cr 0 and Cred
would lead to the same credence function.
The following is an example of a requirement of rationality for Cr 0
(and hence for Cred) which has no analogue for credence functions.
As we shall see later, this requirement leads to one of the most impor-
tant axioms of inductive logic. (The term "individual" means "element
of the universe of discourse", or "element of the population" in the
terminology of statistics.)
R4. Requirement of symmetry. Let at and a, be two distinct
individuals. Let H and H' be two propositions such that H' results
from H by taking a3 for a( and vice versa. Then Cr 0 should be such that
Cr„( H) = Cr 0 (//')- (I n other words, Cr 0 should be invariant with
respect to any finite permutation of individuals.)
This requirement seems indispensable. H and H' have exactly the
same logical form; they differ merely by their reference to two distinct
individuals. These individuals may happen to be quite different. But
since their differences are not known to X at time T0, they cannot have
any influence on the Cr 0 -values of H and H'. Suppose, however, that at
a later time Tn, X's knowledge Kn contains information E relevant to
H and H', say information making H more probable than H' (as an
extreme case, E may imply that H is true and H' is false). Then X's
credence function Cr n at Tn will have different values for H and for H'.
Thus it is clear that R4 applies only to Cr„, but is not generally valid
for other credence functions Cr„ (n > 0).
Suppose that X is a robot constructed by us. Since the propositions
H and H' are alike in all their logical properties, it would be entirely
arbitrary and therefore unreasonable for us to assign to them different
Cr 0 -values in the construction of X.
A function Cr 0 is suitable for being built into a robot only if it
fulfills the requirements of rationality; and most of these requirements
apply only to Cr 0 (and Cred) but not generally to other credence
functions.
6
Inductive Logic
N o w we are ready to take the step from normative decision theory

to inductive logic. This step consists in the transition from the concepts
of a rational Cr„-function and a rational Cred-function to corresponding
purely logical concepts. The former concepts are quasi-psychological;
they are assigned to an imaginary subject X supposed to be equipped
with perfect rationality and an unfailing memory; the logical concepts,
in contrast, have nothing to do with observers and agents, whether
natural or constructed, real or imaginary. For a logical function
corresponding to Cr 0 , I shall use the symbol and I call such func-
tions (inductive) measure functions or ^ - f u n c t i o n s ; for a logical
function corresponding to Cred, I shall use the symbol and I call
these functions (inductive) confirmation functions or ^-functions. I
read i ( € { H | E ) ' as "the degree of confirmation (or briefly "the confirma-
t i o n " ) of H with respect to £ " (or: " . . . given £ " ) . A n ^ - f u n c t i o n is a
function from propositions to real numbers. A n y ^ - f u n c t i o n J i or
^-function ^ is supposed to be defined in a purely logical way, i.e.,
on the basis of concepts of logic (in the wide sense, including set theory
and hence the whole of pure mathematics). Therefore the values of J i
or of for given propositions depend merely on the logical (set-
theoretic) properties of these propositions (which are sets in a prob-
ability space) but not on any contingent facts of nature (e.g., the truth
of the propositions).
Inductive logic studies those ^ - f u n c t i o n s that correspond to
rational Cr 0 -functions, and those ^-functions that correspond to
rational Cred-functions. Suppose jffl is a logically defined ^#-function.
Let us imagine a subject X whose function Cr 0 corresponds to Jt, i.e.,
for every proposition H, Cr 0 (//) = ¿ffl(H). If we find that Cr 0 violates
one of the rationality requirements, say R 4 , then we would reject this
function Cr 0 , say for a robot we plan to build. Consequently we also
exclude the corresponding function J i from those treated as admissible
in the system of inductive logic we plan to construct. Therefore, we set
up axioms of inductive logic about ^ - f u n c t i o n s so that these axioms
correspond to the requirements of rationality which we find in the theory
of rational decision making about Cr 0 -functions.
For example, we may lay down as the basic axioms of inductive
logic for J £ the usual axioms of the calculus of probability. These
axioms correspond to the requirement R1 of coherence, by virtue of
theorem (7). Further we may have an axiom saying that jffl is regular.
This axiom corresponds to the requirement R2 of strict coherence by
theorem (10).
Then we shall have in inductive logic, in analogy to the requirement

R4 of symmetry, the following:
(18) Axiom of Symmetry. </M is invariant with respect to any
finite permutation of individuals.
There is another alternative: we may state axioms for instead
of those for J t . (This method is used in [Art. 2].)
All axioms of inductive logic state relations among values of
Jt or ^ as dependent only upon the logical properties and relations
of the propositions involved (with respect to language systems with
specified logical rules). Inductive logic is the theory based upon these
axioms. It may be regarded as a part of logic in view of the fact that
the concepts occurring are logical concepts. (Exactly speaking, this
holds only for pure inductive logic, not for applied inductive logic. The
distinction between these two fields is discussed in [Art. 2, §4A].) It is
an interesting result that this part of normative decision theory,
namely, the logical theory of the ^ - f u n c t i o n s and the ^-functions, can
thus be separated from the rest. We should note, however, that this
logical theory deals only with the abstract, formal aspects of proba-
bility, and that the full meaning of (personal) probability can be
understood only in the wider context of decision theory through the
connections between probability and the concepts of utility and
rational action.
It is important to notice clearly the following distinction. While
the axioms of inductive logic themselves are formulated in purely
logical terms and do not refer to any contingent matters of fact, the
reasons for our choice of the axioms are not purely logical. For example,
when asked why I accept the axiom of symmetry (18), I would point
out that if X has a Cr„-function corresponding to an ^ - f u n c t i o n
violating (18), then his function Cr 0 would violate R4, and I show that
therefore X, in a certain possible knowledge situation, would be led to
an unreasonable decision. Thus, in order to give my reasons for the
axiom, I move from pure logic to the context of decision theory and
speak about beliefs, actions, possible losses, and the like. However,
these considerations are not in the field of descriptive, but of normative
decision theory. Therefore, in giving my reasons, I do not refer to
particular empirical results concerning particular agents or particular
states of nature and the like. Rather, I refer to a conceivable series of
observations by X, to conceivable sets of possible acts, to possible
states of nature, to possible outcomes of the acts, and the like. These
features are characteristic for an analysis of reasonableness of a given
function Cr 0 , in contrast with an investigation of the successfulness of
the (initial or later) credence function of a given person in the real
world. Success depends upon the particular contingent circumstances,
rationality does not.
INDUCTIVE LOGIC A N D R A T I O N A L DECISIONS 27
There is a class of axioms of inductive logic which I call axioms of

invariance. The axiom of symmetry is one of them. Another axiom says
that, under certain conditions, ¿ffl is invariant with respect to any finite
permutation of attributes belonging to a family of attributes, e.g.,
colors, provided these attributes are alike in their logical properties.
These and other invariance axioms may be regarded as representing the
valid core of the old principle of indifference (or principle of insufficient
reason). The principle in its original form, as used by Laplace and other
authors in the classical period of the theory of probability, was certainly
too strong. It was later correctly criticized by showing that it led to
absurd results.2 I believe, however, that the basic idea of the principle
is sound. Our task is to restate it by specific restricted axioms.
It seems that for personal probability most authors do not accept
any axioms of invariance. In the case of those authors who take credence
as their basic concept, e.g., Ramsey, de Finetti, and Savage, this is
inevitable, since the invariance axioms do not generally hold for
credence functions. In order to obtain a stronger system, it is necessary
to take as the basic concept either initial credence or credibility (or
other concepts in terms of which these are definable).
When we construct an axiom system for J i , then the addition of
each new axiom has the effect of excluding certain ^"-functions. W e
accept an axiom if we recognize that the ^ - f u n c t i o n s excluded by it
correspond to nonrational Cr 0 -functions. Even on the basis of all
axioms that I would accept at the present time for a simple qualitative
language (with one-place predicates only, without physical magnitudes),
the number of admissible ^ - f u n c t i o n s , i.e., those that satisfy all
accepted axioms, is still infinite; but their class is immensely smaller
than that of all coherent ^ - f u n c t i o n s . There will presumably be further
axioms, justified in the same way by considerations of rationality. W e
do not know today whether in this future development the number of
admissible ^ - f u n c t i o n s will always remain infinite or will become
finite and possibly even be reduced to one. Therefore, at the present
time I do not assert that there is only one rational Cr 0 -function.
2 On this question, compare my [19536,pp. 193 f . ] reprinted in [1955,pp.21 f.].

7
The Question of Acceptance
I think that the theory of the J i - and ^-functions deserves the

often misused name of "inductive logic". Earlier I gave my reasons for
regarding this theory as a part of logic. The term "inductive" seems
appropriate because this theory provides the foundation for inductive
reasoning (in a wide sense). I agree in this view with John Maynard
Keynes and Harold Jeffreys. It is important, however, that we recognize
clearly the essential form of inductive reasoning. My view on this point
differs from that of almost all writers on induction in the past and the
great majority of contemporary writers. They regard inductive reason-
ing as an inference leading from some known propositions, called the
premisses or evidence, to a new proposition, called the conclusion,
usually a law or a singular prediction. From this point of view the result
of any particular inductive reasoning is the acceptance of a new propo-
sition (or its rejection, or its suspension until further evidence is found,
as the case may be). I have serious doubts about this view. If we hold it,
we are unable to refute Hume's dictum that there are no rational
reasons for induction. Suppose that we find in earlier weather reports
that a weather situation like the one we have today has occurred one
hundred times and that it was followed each time by rain the next
morning. (If someone objects that the number one hundred is too small,
let him take one hundred thousand or any number he regards as
sufficiently large.) According to the customary view, on the basis of
this evidence the "inductive method" entitles us to accept the prediction
that it will rain tomorrow morning. But then Hume is certainly right
in protesting that we have no rational reason for the acceptance of this
prediction, since, as everyone will agree, it is still possible that it will
not rain tomorrow.
In contrast with this view, it seems to me that the result of a
paradigmatic piece of inductive reasoning with respect to a hypothesis
H, starting from the evidence E, consists in an assignment of a prob-
ability to H, namely the value of H | E).
I do not regard it as generally wrong to accept or reject propositions.
For the purpose of a special investigation, an investigator will usually
accept some general assumptions (see my discussion of basic assump-
tions in [Art. 2, §5]), or specific assumptions, for example about the
reliability of his measuring instruments. But I would not say that
such assumptions are conclusions obtained by "inductive inference".
For further discussion on acceptance compare my discussion remarks
"On rules of acceptance" [1968, pp. 146 ff.] and Jeffrey [1956],
30 STUDIES IN INDUCTIVE LOGIC A N D PROBABILITY
Let us apply this new view of the essential task of inductive reason-
ing to the earlier example. Let H be the prediction of rain for tomorrow.
According to the new view, X does not assert the prediction H, but
merely the following statements (where the value 0.99 serves simply as
an example):
(19) (a) At the present moment Tn, the totality of A^s observation
results is Kn.
(b) <g{H | Kn) = 0.99.

(c) Cred | Kn) = 0.99.
(d) Crx Tn(H) = 0.99.
(a) is the statement of the evidence at hand, the same as in the

first case. But now, instead of accepting H, X asserts the statement (c)
of the Cred-value for H on his evidence, (c) is the result of A"s inductive
reasoning. Against this result Hume's objection does not hold, because
X can give rational reasons for it. (c) is derived from (b) because X
has chosen a mathematically defined function ^ as his credibility func-
tion. (b) is an analytic statement based on the definition of <«?. A"s
choice of if was guided (though not uniquely determined) by the axioms
of inductive logic. And for each of the axioms we can give a reason,
namely, a rationality requirement for credibility functions. Thus ^
represents a reasonable credibility function. Finally, A"s credence
value (d) is derived from (c) and (a) by (15).
Now some philosophers, including some of my empiricist friends,
would raise the following objection. If the result of inductive reasoning
is merely an analytic statement (like (b) or (c)), then induction cannot
fulfill the task of guiding our practical decisions. As a basis for a
decision we need a statement with factual content. If we do not wish
to accept the prediction H itself, then, they say, we must use a statement
of the statistical probability of H, which is a factual statement. In
answer to this objection I would first point out that X does have a
factual basis in his evidence, as stated in (a). And for the determination
of a rational decision neither the acceptance of H nor knowledge of the
statistical probability of H is needed. The rational personal probability,
i.e., the credence as stated in (d), is sufficient for determining first the
rational value of each possible act by (16), and then a rational decision.
Thus, in our example, in view of (b), X would decide to make a bet on
rain tomorrow if it were offered to him at odds of ninety-nine to one or
less, but not more.
The old puzzle of induction consists in the following dilemma.
On the one hand we see that inductive reasoning is used by the scientist
and the man in the street every day without apparent scruples; and we
have the feeling that it is valid and indispensable. On the other hand,
once Hume awakens our intellectual conscience, we recognize that

here is a serious difficulty. Who is right, the man of common sense or
the critical philosopher? We see here, as so often, that both are partially
right. Hume's criticism of the customary forms of induction was correct.
But still the basic idea of common sense thinking is vindicated: induc-
tion, if properly reformulated, can be shown to be valid by rational
criteria.
2
A Basic System
of Inductive Logic
Parti
BY R U D O L F C A R N A P
1
Basic Concepts and Basic Axioms
A. BASIC CONCEPTS
"Inductive logic" is here understood as the theory of probability

in the logical or inductive sense, in distinction to probability in the sta-
tistical sense, measured by frequencies. This and the subsequent articles
are based on the conception of the nature of inductive logic as explained
in [Art. 1] (i.e., Article 1 in this volume) and treated in greater detail in
my book [1950] (see the Bibliography at the end of this volume).
(Knowledge of the book is not presupposed in the present article.)
In my book the functions m (measure function, absolute probabil-
ity) and c (confirmation function, relative probability) were applied to
sentences in some language. However, in the case of more complex
conceptual systems, e.g., those involving real numbers and real-valued
functions, no language can express all possible cases by sentences (or
even by classes of sentences). Therefore, I have here chosen to take as
arguments of the probability functions not sentences, but events or
propositions. I use the latter two terms as synonymous. In probability
theory the term "event" is more customary, in logic the term
"proposition". The term "event" has here, of course, a much wider
sense than in ordinary language. It refers not only to actual events, but
also to merely possible events. Further, in our sense, the series of all
eclipses of the moon is one event; likewise Newton's law of gravitation,
no matter whether true or false. And there is even one impossible
event. (We shall, however, not use formulations like "two successive
occurrences of the same event"; we shall say instead "two successive
events of the same kind" or "two successive individuals having the same
attribute".) The term "proposition" is often used in logic for certain
intensional entities, regarded with suspicion by some philosophers.
Here, however, we shall represent propositions in an extensional way
(namely, by sets of models, §3B).
Instead of the functions m and c applied to sentences we shall use
corresponding functions and <€ applied to propositions. Nevertheless,
each discussion of inductive logic in these articles will refer to some
language or class of languages, and in particular to the set of descriptive
(i.e., nonlogical) constants of the language. For any language J ? ,
is the class of events or propositions on . £ x contains all events
described by sentences in but in general it contains also others. For
any language form that we shall discuss, we shall define the correspond-
ing class £ ( W e shall often omit the subscript "jSf" in these and similar
notations, if only one language is talked about.)
We use 'H\ 'A', ' 5 ' , etc., for propositions. Thus '<g(H\Ey
takes the place of the earlier 'c(A, e)\ where h and e were sentences.
Since propositions will be defined as certain sets, set connectives will
be used instead of sentence connectives. Thus the complement — E
corresponds to the negation the intersection E n H corresponds
to the conjunction e. h, and the union E U H corresponds to the
disjunction e v h. The necessary proposition Z (the set of all models, §3)
corresponds to any tautological sentence; the empty (or impossible)
proposition 0 corresponds to any self-contradictory sentence. Inclu-
sion, e.g., E c H, corresponds to logical implication, e.g., V e => h.
Other set-theoretic notations: {x} is the set whose only element is
x; analogously, {x,y}, {x l 5 . . . , xn}. {x: . . . x . . .} is the set A such
that x e A if-if x satisfies the condition . . . . v . . . . (I write 'if-if' for
'if and only if'.) If si is a class of sets, U si is the union of si, p| is
the intersection of si. (x,y) is the ordered pair with x as the first
member and y as the second; analogously, (x,y, z) is an ordered
triple, (x 1; . . . , xn) an ordered n-tuple. A \ B = D f {(x, j ) : x e A and
yeB); the Cartesian product A X B is the set of all ordered pairs
whose first member belongs to A and whose second member to B.
We shall use the terms "class" and "set" as synonyms. We
sometimes, however, prefer "class of sets" to "set of sets" (and
therefore also "class of functions", since functions may be regarded
as sets of pairs or of n-tuples).
We use 'I' for the set of all integers; 'N' for the set of the natural
numbers (the positive integers), and °N for N u {0}. A set A is said to
be (¡enumerable if-if it has the same cardinal number as N (namely,
aleph-zero). A is said to be countable if-if A is finite or denumerable.
Dl-1. a. Two sets A, B are disjoint = D ( A and B have no common

element (in other words, A n B = 0).
b. si is a disjoint class of sets = D£ any two distinct sets
A, B e si are disjoint.
c. The class si is a covering of the set A =D£ A c y .
d. The class (or sequence) si is a partition of the set A =ms/
is a disjoint class of one or more sets, and (J si = A.
Dl-2. Let X be a nonempty set and si be a class of subsets of X.

a. si is a field on X = D£
(i) J e r f ;
(ii) if A esi, X - A g si;
(iii) if A, B e si, A U Be si.
b. si is a cr-field on X =D£ si is a field on X, and for any
countable subclass ¿8 of si, (J e si.
A BASIC SYSTEM OF INDUCTIVE LOGIC, PART I 37
A field is closed under all finite sequences of Boolean operations (union,

intersection, difference). A cr-field is closed under all countable
sequences of such operations.
The intersection of any class of fields [cr-fields] on X is again a
field [cr-field] on X. We define:
Dl-3. Let X be a nonempty set, and si be a class of subsets of X.

38 is (a) the field [(b) the cr-field] on X generated by si =Df 38 is the
intersection of all those fields [cr-fields] on X which include si.
Tl-1. Let stf be any class of subsets of X, and 88 be (a) the field
[(b) the cr-field] on X generated by si. Then 38 is the only class that
satisfies the following three conditions:
a. si ^ 38,
b. 38 is a field [a cr-field] on X,
c. 38 is the smallest class satisfying (a) and (b); that is to say,
if any class J " satisfies (a) and (b), then 38 <=- 38'.
We shall assume that for any language , the class of sentences in

and the class of propositions § s a t i s f y the following conditions.
(1-1) (i) The class of sentences is closed with respect to the following
sentence-forming operations: negation, disjunction, and
conjunction.
(ii) For every sentence S in JSf, there is a proposition E g e S
corresponding to S; we say then also that S designates or
describes Es.
(iii) $ is a a-field.
For any language JSf to be used we shall proceed as follows. We

shall specify a certain class of functions, called the model-functions
or simply the models of =5?. (The models are the "points" of our
"probability space".) £ will be defined as a certain cr-field on Z. In
most cases we shall define first a subclass of § , namely <^bas, the class
of the basic propositions on (sometimes defined as the propositions
corresponding to certain sentences, called the basic sentences of JSf).
Then we shall define & as the cr-field generated by <?ba3 on Z.
Each model represents a most specific possible case with respect
to y . The two extreme propositions are the necessary proposition,
holding in every possible case, and the empty event or the impossible
proposition, holding in none of the possible cases. We define these and
related concepts as follows:
Dl-4. Let E and E' be propositions (events).

a. E is necessary = D£ E = Z.
b. E is impossible (or empty) = D£ E = 0.
c. E is possible (or nonempty) —mE^ 0.
d. E is contingent (or factual) = D£ E is neither necessary nor
impossible.
e. E is included in E' (or E strictly implies £") = D£ E <= E'.
B. THE BASIC AXIOMS

In inductive logic, the "^-functions are more important than the
^"-functions, since some ^-values represent rational degrees of belief
and thus help in determining rational decisions. The ^ - f u n c t i o n s
serve mainly as convenient means for defining ^-functions and for
determining their values. An ^ - f u n c t i o n is an (absolute) probability
function in the sense of the axiomatic probability calculus (which is
neutral with respect to interpretations); a "^-function is a conditional
(or relative) probability function. In some systems, J i is taken as
primitive and ^ is defined on its basis. Some ^-functions, however,
cannot be defined on the basis of corresponding ^ - f u n c t i o n s by the
customary definition in form of a quotient (e.g., x<€ for 1 = 0, my
[1952a] §14]). Therefore we take here as primitive and J t as defined,
and we state axioms only for <€. In the following it is always assumed
that the second argument of # is nonempty.
Al. Axiom of the lower bound: | E) ^ 0.

A2. Axiom of self-confirmation: ^(E | E) = 1.
A3. Axiom of the complement: '£(H\E) + <g{-H\E) = 1.
A4. General multiplication principle: If E n H is possible,
V(H C\ H'\E) = |Eftifl' |E r\H).
The first three axioms are here given in forms weaker than the
customary ones (Al instead of a combination of Al with the subsequent
theorem T2a; A2 instead of T2b; A3 instead of the special addition
principle T2h). Von Wright ([1957], see the Bibliography at the end
of this volume) proposed these weaker forms and showed that the
customary forms are provable on this basis. The end of a proof is
marked by ' • ' .
Tl-2. Some consequences of the axioms.

a. <<o{H | £ • ) < ! . (The principle of the upper bound.)
A BASIC SYSTEM OF INDUCTIVE LOGIC, PART I 39
Proof. From A3: | £ ) = 1 - ^ { - H \ £).

From A1: ^ ( — H \ £ ) ^ 0. Hence the assertion. •
b. If £ c H, <V(H\E) = 1.
Proof. Let E c H. Then E n H = E. Hence
«"(£ | E) = V(E C\H\E) = <#{E | E)<£(H \ E) (A4).
Hence the assertion with A2. •
c. If E -H, <g(H | E) = 0. (From (b) and A3.)

d. %(Z\E) = 1. (From (b).)
e. <^(0 | E) = 0. (From (c).)
f. <#{H | E) = C\H'\E)+ <<g{H n —H' | E).
Proof. 1. If E r\H= 0 , this follows from (c).—2. Let E O H # 0 .
Then from A3: \ E n if) + \ E n H) = 1. By multiply-
ing with ^ ( H [ £ ) on both sides and exchanging the sides, we obtain:
V(H | £) = <T(/i | £)*f(J¥' | £ n tf) + | E)<#(-H' |£ n #).
Transforming both products by A4 yields the assertion. •
g. if U H' | £) = | £) + " g W | £) - <£{H O / T | £).
Proof. From (f):

^ ( / f U i f ' | £ ) = i?((if U 7/') O H' I £) + ^((Zf U H') n I £)
= « x / r | £) + n - / r | £).
Again from (f): C(H n | £) = i?(if | £) - n | £).
Hence the assertion. •
+h. Special addition principle. If £ n H n H' = 0 ,

^ ( t f u i f ' I £ ) = «"(Ji I £ ) + ^ ( / Z ' I £).
Proof. Under the condition stated, ^ ( 7 / n / / ' | £ ) = 0 (from (c)).

Hence the assertion with (g). •
Thus the four axioms are equivalent to most of the customary

axiom systems for conditional probability. Some systems contain
another axiom, which we shall give later in a restricted form (the axiom
of regularity, A5). [For c with sentences as arguments, another axiom is
necessary to effect that the value of c(h, e) remains unchanged if either
h or e is replaced by a logically equivalent sentence.]
Another random document with
no related content on Scribd:
d’un rose livide, qui pâlissaient et s’avivaient tout à tour, la
tourmentant de démangeaisons douloureuses.
Elle n’y prit garde, s’en cacha de son amant par une sorte de
pudeur, et aussi pour ne pas lui créer des inquiétudes inutiles. Elle
ne s’en alarma pas elle-même, mettant le tout sur le compte des
fatigues de leur voyage, de leur nouvelle installation et de ses tristes
émotions encore récentes.
Mais peu après, Hamidou se trouva comme elle, éprouvant ce
qu’elle éprouvait, et alors, s’étant mutuellement renseignés, tous
deux prirent le parti d’en rire.
Ils en riaient d’autant plus qu’ils crurent avoir une scarlatine
légère, selon le diagnostic d’un vieux sorcier moghrebin, leur voisin,
lequel soignait, pour ce mal-là, quelques enfants de la ville arabe.
Et de fait, sans autre traitement que quelques simples anodins
ordonnés par le vieux marabout et accompagnés de quelques
versets du Coran, le mal disparut et ne laissa pas la moindre trace.
Ce furent, alors, pour les deux amants, des mois d’un bonheur
parfait, inouï, et comme le Rétributeur n’en concède que très
rarement à des créatures élues, dont il marqua le front du doigt pour
des desseins pleins de mystère.
Le merveilleux talent d’Hamidou était apprécié du Mozabite ; et

les salaires qu’il gagnait suffisaient pour assurer leur existence, au
delà même, car ils vivaient simplement de la vie arabe, comme
d’ailleurs tous leurs voisins, dans ce quartier musulman de la ville
haute.
D’abord, par prudence, comme on l’a vu, puis par goût et par
habitude, enfin et surtout pour complaire à son amant, qui la
préférait ainsi, la Madalena s’habillait comme ses voisines.
Elle se voilait comme elles pour sortir, et, comme elles, sortait
rarement pour les besoins du ménage ou pour aller au bain maure,
accompagnée d’une vieille négresse qu’Hamidou lui avait donnée
pour servante.
Enfin, pour combler leur amoureuse béatitude, Hamidou obtint de
son Mozabite qu’il emporterait et ferait chez lui son ouvrage.
Dès lors, ils ne se quittèrent plus un instant, et la claustration à
laquelle ils se vouèrent d’un accord tacite, eut, pour eux, des
voluptés ineffables.
Bientôt, de la Sicilienne, il ne resta plus grand’chose en la
Madalena. Conquise par l’amour à l’Islam dont sa grande âme avait
compris, depuis longtemps, l’austère et simple beauté, ce fut avec
un bonheur profond que, pour donner à son amant cette joie
suprême, elle prononça, devant le vieux marabout moghrebin, les
paroles qui la sacrèrent musulmane :
« Il n’y a d’autre Dieu qu’Allah et Mohammed est son Prophète ».
N’ayant qu’une âme, pouvaient-ils, vraiment, ne pas avoir même
idéal, même Dieu, mêmes espoirs, de s’aimer encore et toujours
dans le même ciel que leur ouvriraient les mêmes prières ?
L’Islam, avec le bel épanouissement qu’il donne à la vie
intérieure, n’a-t-il pas d’ailleurs, pour ceux qui s’aiment,
d’incomparables délices ?
Désormais donc, la Madalena devint vraiment ce que ses
voisines et voisins l’avaient toujours crue : Lella Zina, une pieuse
Moghrebine amenée du pays natal par son époux et qui, par sa
bonté, ne tardait pas à se faire adorer de tout le monde.
L’aisance, en effet, nécessaire à la charité, venait chaque jour,
car ils étaient deux, maintenant, à exécuter des broderies
merveilleuses. L’amour ne fut-il pas, de tout temps, le plus puissant
des thaumaturges ? Quelques mois lui avaient suffi pour faire de la
Madalena une brodeuse sur étoffe, comme Hamidou, incomparable.
Une part de l’argent qu’elle gagnait, servait, d’un commun accord, à
secourir autour d’eux les plus criantes misères.
Et Dieu sait si, dans cette pauvre Kasbah, elles sont
nombreuses.
Enfin, pour tout le quartier musulman, elle était bien Lella, la
Sainte, la Madame, autant dire une Maraboute.
Quand elle sortait pour aller, enveloppée et alourdie de ses
voiles, à travers les ruelles de la Kasbah, les enfants se disputaient
pour avoir d’elle une caresse, les vieillards, une main au cœur,
inclinaient la tête, et les pauvresses qu’elle soulageait pleuraient en
baisant le pan de son bel haïk de soie blanche.
Et autant se réjouissait Hamidou de la voir ainsi aimée, vénérée
et sainte que de la sentir belle et sienne.
*
* *
Or, voici, qu’au plein de cette félicité pareille à nulle autre, de

nouveau la Madalena se sentit malade : mais, cette fois, au grand
désespoir d’Hamidou, son mal s’aggrava avec une rapidité terrible.
Ce furent, d’abord, des douleurs qui, pendant la nuit, lui broyaient le
crâne, comme si on l’eût martelé sur une enclume. Elle se réveillait
brusquement en poussant des cris dont s’épouvantait Hamidou qui
ne savait que lui faire.
Le vieux Moghrebin consulté, ne s’alarma pas, disant que ça
passerait comme le reste. Il écrivit sur des petits carrés de papier,
quelques sourates du Saint-Livre, en fit des boulettes qu’elle devait
avaler comme pilules, et déclara qu’Allah, miséricordieux et clément,
se chargeait du reste.
Mais hélas ! Il fit si peu que, loin de s’atténuer, les douleurs du
crâne redoublèrent et s’étendirent même à tous les os de la malade,
lui arrachant, pendant la nuit, des cris qu’elle ne pouvait réprimer,
malgré toute sa volonté de ne pas ajouter à la détresse du pauvre
Hamidou couché près d’elle.
Pourtant, ce qui la désespérait plus encore que ces tortures
nocturnes, c’était de se voir maigrir, pour ainsi dire, à vue d’œil. De
plus, chaque matin, quand elle démêlait devant son miroir sa
splendide chevelure, il en restait, aux dents de son peigne, des
poignées énormes, qui allaient de jour en jour grossissant, et dont la
vue mouillait son front d’une sueur froide. Enfin, après chacune de
ces crises aiguës, elle sentait branler et voyait jaunir ses dents,
jusqu’alors plus blanches et plus éclatantes que des perles.
Et ce lent évanouissement de sa beauté, mais par-dessus tout, la
perte bientôt consommée de sa chevelure, la jetait en de mornes
désespoirs qu’elle s’efforçait, sans y parvenir, de dissimuler à
Hamidou, dont la navrance était encore plus profonde que la sienne.
Il se résolut donc, enfin, ce par quoi il aurait dû commencer, à
faire appeler un des médecins de la ville, mais, avant, il perdit
encore un temps précieux en mandant, sur le conseil d’une voisine,
la vieille Frendah, cette sorcière de Bab-el-Oued, dont la réputation
de guérisseuse dépassait le Sahel d’Alger, et que, de leurs très
lointains douars du Sud, accouraient consulter de riches nomades.
Elle arriva, palpa longuement le crâne de la malade, et tous les
os de ses membres, et annonça que c’était un djin amoureux de
Lella-Zina qui, chaque nuit, venait lui frapper la tête et les membres,
et lui arracher un peu de sa chevelure.
Elle se livra à de bizarres incantations, ordonna de faire boire à la
malade une mixture plus étrange encore, dans laquelle entraient du
lait d’ânesse, de l’urine d’âne, des pépins d’orange et quelques
graines de courge, pilées ensemble.
Elle ordonna, en outre, de lui appliquer une peau de chèvre noire
sur la poitrine et sur les membres.
Le mal bien entendu empira : les beaux cheveux de Lella Zina
restèrent plus encore à son peigne ; quelques dents achevant de se
déchausser, tombèrent ; son amaigrissement s’accrut, la peau de
ses bras devint squameuse comme celle des couleuvres, et il lui
sortit, aux doigts des pieds et des mains, tout autour des ongles
qu’elles firent lentement tomber, de petites plaies douloureuses et
très profondes.
Enfin, quand le docteur arriva, il n’eut pas de peine à reconnaître
la nature du terrible mal, et put, par une médication énergique, en
arrêter la marche ascendante, mais la science, déclara-t-il, était
impuissante devant les ravages accomplis et qui faisaient de la belle
Lella Zina une créature méconnaissable.
Alors, à certaines paroles et investigations du médecin, la pauvre
femme comprit, elle aussi, quel était son mal, et eut de lamentables
intuitions sur son origine.
Elle se rappela, en effet, que pendant un de ces accès de rage
jalouse, Vittorio lui avait dit au milieu d’un tas d’injures immondes :
— Un jour, tu seras pareille à Thérésa la Gouge.
Elle frémit, mais courba la tête, résignée comme une bonne
musulmane.
Pourtant, elle se sentit incapable de survivre à cette beauté qui
lui valut l’amour d’Hamidou, et était sa seule raison d’être en ce
monde.
— Ma vie ne peut désormais être pour lui qu’une continuelle
déplaisance, il ne me reste donc qu’à mourir, pensait-elle, tandis
qu’Hamidou, refoulant au fond de son âme bonne, sa désespérance,
redoublait de dévouement, de tendresse délicate, de tendresse
d’amant épris, pour lui rendre l’existence encore meilleure.
— Eh ! que t’importe, trésor, ne cessait-il de lui répéter, puisque
tu es toujours belle à mes yeux et que je t’aime ! »
Or, comme il avait eu l’affectueuse précaution de faire enlever du
logis toutes les glaces, la Madalena ne pouvant plus se regarder et
entendant ces douces paroles sentait s’apaiser son ardent désir de
suicide.
Mais un jour, ayant mis la main sur un de ces petits miroirs
engaînés de cuir brodé dont se servent les mauresques, elle eut la
triste curiosité de s’y contempler, se vit, et poussa un rugissement de
détresse, tant elle s’apparut semblable à Thérésa la Gouge.
Hamidou était derrière elle, et lui enlevait, mais trop tard, la fatale
glace.
La pâleur de la pauvre femme ajoutait encore à la laideur de son
visage. Alors, n’osant plus même l’embrasser, comme une chienne
ou une esclave, elle se jeta à ses pieds, les baisa dans une étreinte
frénétique.
— Hamidou, clama-t-elle, de grâce, ne m’empêche plus de
mourir. Ecoute, tu m’as donné la plus grande preuve d’amour qu’un
amant puisse donner à son amante. Tu m’as aimée laide et malade
autant, même plus, que dans la splendeur de ma beauté et de ma
saine jeunesse. Nulle créature n’a donc eu plus de bonheur que ta
servante. Oh ! mourir, maintenant, encore tiède de tes caresses !
ami, crois-moi, ce suprême sourire de ma Destinée serait plus doux
que les autres. La tienne te réserve encore des matins roses, car tu
es jeune et tu es beau, et tu mérites d’être aimé jusqu’à la tombe.
Ami, ami, je t’en supplie, laisse-moi mourir en baisant ta main
secourable et bonne.
Devant ces accents de désespoir, Hamidou vit bien que tout était
fini, que tout bonheur était désormais pour elle impossible. Il aurait
beau l’aimer ardemment dans sa beauté évanouie, se dépenser en
mille efforts pour lui faire oublier la cruauté du destin, l’image de sa
laideur ne cesserait de la hanter et empoisonnerait toutes les heures
de son existence.
Et lui, Hamidou, pourrait-il encore être heureux en la sentant
malheureuse ? Alors, il regretta d’être épargné par le mal, et les
paroles du sorcier de Bab-Ménara lui revinrent à la mémoire, avec
une netteté sans pareille :
— Une roumie prendra ton cœur, et pour le reste, permets au
vieil Abdallah de clore ses lèvres.
Ce reste, il le connaissait à cette heure.
Il courba la tête devant l’inéluctable Mektoub, et serrant
tendrement dans ses bras son amante :
— Trésor, fit-il simplement, la vie sans toi serait pour moi pire que
la mort, et puisque tel est notre destin, mourons ensemble.
Et le soir même, ils s’enfermèrent dans leur chambre, allumèrent
le fatal réchaud, puis ayant revêtu leurs plus beaux habits,
s’endormirent dans la paix de Dieu, après un baiser suprême.
Le lendemain, la vieille négresse qui couchait sur une natte dans
la cour, ne les voyant pas sortir, força la porte et les trouva morts, les
mains enlacées et se souriant encore l’un à l’autre.
Elle comprit tout le drame, l’approuva dans son âme simple, et
afin que rien ne vînt ternir, aux yeux des musulmans, la réputation
de sainteté dont jouissait sa bonne maîtresse, elle enleva le
réchaud, et après s’être assurée que rien ne pouvait trahir le suicide,
elle appela les voisins à l’aide.
Ils accoururent avec le vieux moghrebin en tête, et tous furent
étonnés du sourire de béatitude qui voltigeait sur les lèvres des deux
cadavres.
— « Mektoub ! » murmura le marabout après avoir constaté la
mort ; et il déclara que, sur l’ordre même d’Allah, l’ange Azrael avait
dû leur apparaître dans le jardin des délices. Ce qui expliquait la
douceur ineffable de leur sourire.
Quand on voulut les séparer pour remettre leur dépouille à celui
et à celle qui lavent les morts, selon les rites, il fut impossible de
rompre l’étreinte qui liait la main droite d’Hamidou à la main gauche
de son amante. Il eût fallu les rompre aux poignets d’un coup de
hache.
Force fut donc — ce que leur conseilla d’ailleurs le vieux
marabout — de les coudre dans le même linceul, et de les coucher
dans la même fosse, leur regard encore brillant d’amour tourné vers
La Mecque.
On leur éleva, dans le cimetière d’El-Kettar, une petite koubba
maraboutique ; et le vendredi qui est le dimanche des musulmans,
les femmes arabes de la Kasbah, viennent, accompagnées de leurs
enfants, prier et manger des confitures parfumées sur la tombe de
Lella Zina qui fut, pendant sa trop courte vie, si secourable et si
bonne.
Isabelle Eberhardt.
SORTI DES PRESSES

DE LA
MAISON EUGÈNE FIGUIÈRE ET Cie
7, RUE CORNEILLE, PARIS
ET
72, RUE VAN ARTEVELDE, BRUXELLES,
LE 15 AVRIL 1913
*** END OF THE PROJECT GUTENBERG EBOOK ISABELLE
EBERHARDT, OU, LA BONNE NOMADE: D'APRÈS DES
DOCUMENTS INÉDITS ***
Updated editions will replace the previous one—the old editions

will be renamed.
Creating the works from print editions not protected by U.S.

copyright law means that no one owns a United States copyright
in these works, so the Foundation (and you!) can copy and
distribute it in the United States without permission and without
paying copyright royalties. Special rules, set forth in the General
Terms of Use part of this license, apply to copying and
distributing Project Gutenberg™ electronic works to protect the
PROJECT GUTENBERG™ concept and trademark. Project
Gutenberg is a registered trademark, and may not be used if
you charge for an eBook, except by following the terms of the
trademark license, including paying royalties for use of the
Project Gutenberg trademark. If you do not charge anything for
copies of this eBook, complying with the trademark license is
very easy. You may use this eBook for nearly any purpose such
as creation of derivative works, reports, performances and
research. Project Gutenberg eBooks may be modified and
printed and given away—you may do practically ANYTHING in
the United States with eBooks not protected by U.S. copyright
law. Redistribution is subject to the trademark license, especially
commercial redistribution.
START: FULL LICENSE

THE FULL PROJECT GUTENBERG LICENSE
PLEASE READ THIS BEFORE YOU DISTRIBUTE OR USE THIS WORK
To protect the Project Gutenberg™ mission of promoting the

free distribution of electronic works, by using or distributing this
work (or any other work associated in any way with the phrase
“Project Gutenberg”), you agree to comply with all the terms of
the Full Project Gutenberg™ License available with this file or
online at www.gutenberg.org/license.
Section 1. General Terms of Use and

Redistributing Project Gutenberg™
electronic works
1.A. By reading or using any part of this Project Gutenberg™
electronic work, you indicate that you have read, understand,
agree to and accept all the terms of this license and intellectual
property (trademark/copyright) agreement. If you do not agree to
abide by all the terms of this agreement, you must cease using
and return or destroy all copies of Project Gutenberg™
electronic works in your possession. If you paid a fee for
obtaining a copy of or access to a Project Gutenberg™
electronic work and you do not agree to be bound by the terms
of this agreement, you may obtain a refund from the person or
entity to whom you paid the fee as set forth in paragraph 1.E.8.
1.B. “Project Gutenberg” is a registered trademark. It may only

be used on or associated in any way with an electronic work by
people who agree to be bound by the terms of this agreement.
There are a few things that you can do with most Project
Gutenberg™ electronic works even without complying with the
full terms of this agreement. See paragraph 1.C below. There
are a lot of things you can do with Project Gutenberg™
electronic works if you follow the terms of this agreement and
help preserve free future access to Project Gutenberg™
electronic works. See paragraph 1.E below.
1.C. The Project Gutenberg Literary Archive Foundation (“the
Foundation” or PGLAF), owns a compilation copyright in the
collection of Project Gutenberg™ electronic works. Nearly all the
individual works in the collection are in the public domain in the
United States. If an individual work is unprotected by copyright
law in the United States and you are located in the United
States, we do not claim a right to prevent you from copying,
distributing, performing, displaying or creating derivative works
based on the work as long as all references to Project
Gutenberg are removed. Of course, we hope that you will
support the Project Gutenberg™ mission of promoting free
access to electronic works by freely sharing Project
Gutenberg™ works in compliance with the terms of this
agreement for keeping the Project Gutenberg™ name
associated with the work. You can easily comply with the terms
of this agreement by keeping this work in the same format with
its attached full Project Gutenberg™ License when you share it
without charge with others.
1.D. The copyright laws of the place where you are located also
govern what you can do with this work. Copyright laws in most
countries are in a constant state of change. If you are outside
the United States, check the laws of your country in addition to
the terms of this agreement before downloading, copying,
displaying, performing, distributing or creating derivative works
based on this work or any other Project Gutenberg™ work. The
Foundation makes no representations concerning the copyright
status of any work in any country other than the United States.
1.E. Unless you have removed all references to Project

Gutenberg:
1.E.1. The following sentence, with active links to, or other

immediate access to, the full Project Gutenberg™ License must
appear prominently whenever any copy of a Project
Gutenberg™ work (any work on which the phrase “Project
Gutenberg” appears, or with which the phrase “Project
Gutenberg” is associated) is accessed, displayed, performed,
viewed, copied or distributed:
This eBook is for the use of anyone anywhere in the United

States and most other parts of the world at no cost and with
almost no restrictions whatsoever. You may copy it, give it
away or re-use it under the terms of the Project Gutenberg
License included with this eBook or online at
www.gutenberg.org. If you are not located in the United
States, you will have to check the laws of the country where
you are located before using this eBook.
1.E.2. If an individual Project Gutenberg™ electronic work is

derived from texts not protected by U.S. copyright law (does not
contain a notice indicating that it is posted with permission of the
copyright holder), the work can be copied and distributed to
anyone in the United States without paying any fees or charges.
If you are redistributing or providing access to a work with the
phrase “Project Gutenberg” associated with or appearing on the
work, you must comply either with the requirements of
paragraphs 1.E.1 through 1.E.7 or obtain permission for the use
of the work and the Project Gutenberg™ trademark as set forth
in paragraphs 1.E.8 or 1.E.9.
1.E.3. If an individual Project Gutenberg™ electronic work is

posted with the permission of the copyright holder, your use and
distribution must comply with both paragraphs 1.E.1 through
1.E.7 and any additional terms imposed by the copyright holder.
Additional terms will be linked to the Project Gutenberg™
License for all works posted with the permission of the copyright
holder found at the beginning of this work.
1.E.4. Do not unlink or detach or remove the full Project

Gutenberg™ License terms from this work, or any files
containing a part of this work or any other work associated with
Project Gutenberg™.
1.E.5. Do not copy, display, perform, distribute or redistribute
this electronic work, or any part of this electronic work, without
prominently displaying the sentence set forth in paragraph 1.E.1
with active links or immediate access to the full terms of the
Project Gutenberg™ License.
1.E.6. You may convert to and distribute this work in any binary,
compressed, marked up, nonproprietary or proprietary form,
including any word processing or hypertext form. However, if
you provide access to or distribute copies of a Project
Gutenberg™ work in a format other than “Plain Vanilla ASCII” or
other format used in the official version posted on the official
Project Gutenberg™ website (www.gutenberg.org), you must, at
no additional cost, fee or expense to the user, provide a copy, a
means of exporting a copy, or a means of obtaining a copy upon
request, of the work in its original “Plain Vanilla ASCII” or other
form. Any alternate format must include the full Project
Gutenberg™ License as specified in paragraph 1.E.1.
1.E.7. Do not charge a fee for access to, viewing, displaying,

performing, copying or distributing any Project Gutenberg™
works unless you comply with paragraph 1.E.8 or 1.E.9.
1.E.8. You may charge a reasonable fee for copies of or

providing access to or distributing Project Gutenberg™
electronic works provided that:
• You pay a royalty fee of 20% of the gross profits you derive
from the use of Project Gutenberg™ works calculated using
the method you already use to calculate your applicable
taxes. The fee is owed to the owner of the Project
Gutenberg™ trademark, but he has agreed to donate
royalties under this paragraph to the Project Gutenberg
Literary Archive Foundation. Royalty payments must be paid
within 60 days following each date on which you prepare (or
are legally required to prepare) your periodic tax returns.
Royalty payments should be clearly marked as such and
sent to the Project Gutenberg Literary Archive Foundation at
the address specified in Section 4, “Information about
donations to the Project Gutenberg Literary Archive
Foundation.”
• You provide a full refund of any money paid by a user who

notifies you in writing (or by e-mail) within 30 days of receipt
that s/he does not agree to the terms of the full Project
Gutenberg™ License. You must require such a user to return
or destroy all copies of the works possessed in a physical
medium and discontinue all use of and all access to other
copies of Project Gutenberg™ works.
• You provide, in accordance with paragraph 1.F.3, a full

refund of any money paid for a work or a replacement copy,
if a defect in the electronic work is discovered and reported
to you within 90 days of receipt of the work.
• You comply with all other terms of this agreement for free
distribution of Project Gutenberg™ works.
1.E.9. If you wish to charge a fee or distribute a Project

Gutenberg™ electronic work or group of works on different
terms than are set forth in this agreement, you must obtain
permission in writing from the Project Gutenberg Literary
Archive Foundation, the manager of the Project Gutenberg™
trademark. Contact the Foundation as set forth in Section 3
below.
1.F.
1.F.1. Project Gutenberg volunteers and employees expend

considerable effort to identify, do copyright research on,
transcribe and proofread works not protected by U.S. copyright
law in creating the Project Gutenberg™ collection. Despite
these efforts, Project Gutenberg™ electronic works, and the
medium on which they may be stored, may contain “Defects,”
such as, but not limited to, incomplete, inaccurate or corrupt
data, transcription errors, a copyright or other intellectual
property infringement, a defective or damaged disk or other
medium, a computer virus, or computer codes that damage or
cannot be read by your equipment.
1.F.2. LIMITED WARRANTY, DISCLAIMER OF DAMAGES -

Except for the “Right of Replacement or Refund” described in
paragraph 1.F.3, the Project Gutenberg Literary Archive
Foundation, the owner of the Project Gutenberg™ trademark,
and any other party distributing a Project Gutenberg™ electronic
work under this agreement, disclaim all liability to you for
damages, costs and expenses, including legal fees. YOU
AGREE THAT YOU HAVE NO REMEDIES FOR NEGLIGENCE,
STRICT LIABILITY, BREACH OF WARRANTY OR BREACH
OF CONTRACT EXCEPT THOSE PROVIDED IN PARAGRAPH
1.F.3. YOU AGREE THAT THE FOUNDATION, THE
TRADEMARK OWNER, AND ANY DISTRIBUTOR UNDER
THIS AGREEMENT WILL NOT BE LIABLE TO YOU FOR
ACTUAL, DIRECT, INDIRECT, CONSEQUENTIAL, PUNITIVE
OR INCIDENTAL DAMAGES EVEN IF YOU GIVE NOTICE OF
THE POSSIBILITY OF SUCH DAMAGE.
1.F.3. LIMITED RIGHT OF REPLACEMENT OR REFUND - If

you discover a defect in this electronic work within 90 days of
receiving it, you can receive a refund of the money (if any) you
paid for it by sending a written explanation to the person you
received the work from. If you received the work on a physical
medium, you must return the medium with your written
explanation. The person or entity that provided you with the
defective work may elect to provide a replacement copy in lieu
of a refund. If you received the work electronically, the person or
entity providing it to you may choose to give you a second
opportunity to receive the work electronically in lieu of a refund.
If the second copy is also defective, you may demand a refund
in writing without further opportunities to fix the problem.
1.F.4. Except for the limited right of replacement or refund set

forth in paragraph 1.F.3, this work is provided to you ‘AS-IS’,
WITH NO OTHER WARRANTIES OF ANY KIND, EXPRESS
OR IMPLIED, INCLUDING BUT NOT LIMITED TO
WARRANTIES OF MERCHANTABILITY OR FITNESS FOR
ANY PURPOSE.
1.F.5. Some states do not allow disclaimers of certain implied

warranties or the exclusion or limitation of certain types of
damages. If any disclaimer or limitation set forth in this
agreement violates the law of the state applicable to this
agreement, the agreement shall be interpreted to make the
maximum disclaimer or limitation permitted by the applicable
state law. The invalidity or unenforceability of any provision of
this agreement shall not void the remaining provisions.
1.F.6. INDEMNITY - You agree to indemnify and hold the

Foundation, the trademark owner, any agent or employee of the
Foundation, anyone providing copies of Project Gutenberg™
electronic works in accordance with this agreement, and any
volunteers associated with the production, promotion and
distribution of Project Gutenberg™ electronic works, harmless
from all liability, costs and expenses, including legal fees, that
arise directly or indirectly from any of the following which you do
or cause to occur: (a) distribution of this or any Project
Gutenberg™ work, (b) alteration, modification, or additions or
deletions to any Project Gutenberg™ work, and (c) any Defect
you cause.
Section 2. Information about the Mission of

Project Gutenberg™
Project Gutenberg™ is synonymous with the free distribution of
electronic works in formats readable by the widest variety of
computers including obsolete, old, middle-aged and new
computers. It exists because of the efforts of hundreds of
volunteers and donations from people in all walks of life.
Volunteers and financial support to provide volunteers with the
assistance they need are critical to reaching Project
Gutenberg™’s goals and ensuring that the Project Gutenberg™
collection will remain freely available for generations to come. In
2001, the Project Gutenberg Literary Archive Foundation was
created to provide a secure and permanent future for Project
Gutenberg™ and future generations. To learn more about the
Project Gutenberg Literary Archive Foundation and how your
efforts and donations can help, see Sections 3 and 4 and the
Foundation information page at www.gutenberg.org.
Section 3. Information about the Project

Gutenberg Literary Archive Foundation
The Project Gutenberg Literary Archive Foundation is a non-
profit 501(c)(3) educational corporation organized under the
laws of the state of Mississippi and granted tax exempt status by
the Internal Revenue Service. The Foundation’s EIN or federal
tax identification number is 64-6221541. Contributions to the
Project Gutenberg Literary Archive Foundation are tax
deductible to the full extent permitted by U.S. federal laws and
your state’s laws.
The Foundation’s business office is located at 809 North 1500

West, Salt Lake City, UT 84116, (801) 596-1887. Email contact
links and up to date contact information can be found at the
Foundation’s website and official page at
www.gutenberg.org/contact
Section 4. Information about Donations to

the Project Gutenberg Literary Archive
Foundation
Project Gutenberg™ depends upon and cannot survive without
widespread public support and donations to carry out its mission
of increasing the number of public domain and licensed works
that can be freely distributed in machine-readable form
accessible by the widest array of equipment including outdated
equipment. Many small donations ($1 to $5,000) are particularly
important to maintaining tax exempt status with the IRS.
The Foundation is committed to complying with the laws

regulating charities and charitable donations in all 50 states of
the United States. Compliance requirements are not uniform
and it takes a considerable effort, much paperwork and many
fees to meet and keep up with these requirements. We do not
solicit donations in locations where we have not received written
confirmation of compliance. To SEND DONATIONS or
determine the status of compliance for any particular state visit
www.gutenberg.org/donate.
While we cannot and do not solicit contributions from states

where we have not met the solicitation requirements, we know
of no prohibition against accepting unsolicited donations from
donors in such states who approach us with offers to donate.
International donations are gratefully accepted, but we cannot

make any statements concerning tax treatment of donations
received from outside the United States. U.S. laws alone swamp
our small staff.
Please check the Project Gutenberg web pages for current

donation methods and addresses. Donations are accepted in a
number of other ways including checks, online payments and
credit card donations. To donate, please visit:
www.gutenberg.org/donate.
Section 5. General Information About Project

Gutenberg™ electronic works
Professor Michael S. Hart was the originator of the Project
Gutenberg™ concept of a library of electronic works that could
be freely shared with anyone. For forty years, he produced and
distributed Project Gutenberg™ eBooks with only a loose
network of volunteer support.
Project Gutenberg™ eBooks are often created from several

printed editions, all of which are confirmed as not protected by
copyright in the U.S. unless a copyright notice is included. Thus,
we do not necessarily keep eBooks in compliance with any
particular paper edition.
Most people start at our website which has the main PG search
facility: www.gutenberg.org.
This website includes information about Project Gutenberg™,

including how to make donations to the Project Gutenberg
Literary Archive Foundation, how to help produce our new
eBooks, and how to subscribe to our email newsletter to hear
about new eBooks.

Studies in Inductive Logic and Probability Volume 1 Rudolf Carnap Editor Online Ebook Texxtbook Full Chapter PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Studies in Inductive Logic and Probability Volume 1 Rudolf Carnap Editor Online Ebook Texxtbook Full Chapter PDF

Uploaded by

Copyright:

Available Formats

Studies in Inductive Logic and

Probability Volume 1 Rudolf Carnap

Studies in Inductive Logic and Probability Volume 2

40 Paradoxes in Logic Probability and Game Theory Presh

Studies in Medieval Philosophy Science and Logic Ernest

American Studies in Europe, Volume 1: Their History and

California Studies in Classical Antiquity Volume 8

California Studies in Classical Antiquity Volume 7

Christianity and the Eastern Slavs Volume 1 California

Selected Collected Works Volume 2 Probability Theory

Rudolf Carnap and Richard C. Jeffrey

University of California Press, Ltd.

Copyright © 1971 by The Regents of the University of California

Library of Congress Catalog Card Number: 77-136025

1. Inductive Logic and Rational Decisions by Rudolf Carnap 5

2. A Basic System of Inductive Logic, Part I by Rudolf Carnap 33

1. Basic Concepts and Basic Axioms 35

11. Representative Functions for J t 131

3. Probability Measures and Integrals by Richard C. Jeffrey 167

4. The Principle of Instantial Relevance by Jürgen Humburg 225

5. Applications of De Finetti's Theorem to Inductive Logic by Haim

Postscript Concerning Extension of Probability Functions 246

Selected Bibliography 253

Carnap's Logical Foundations of Probability [1950] was planned as

work projected in 1950, in the form of a nonperiodical journal in which

Thost-Carnap, for keeping her father's working papers together and

Further Material for the Introduction

1952-1954: Carnap was at the Institute for Advanced Study, in

Inductive Logic and

By "inductive logic" I understand a theory of logical probability

0 m , „ of act Am in the various possible states fV„. [If the situation is

statistical probability are in general not known to X. Therefore this con-

Let us first discuss descriptive decision theory, the theory of

highest betting quotient at which X is willing to bet on H. (As is

Now we go from descriptive to normative decision theory. The

the concept of logical implication in deductive logic. (This logical

Following Shimony [1955], we introduce now a concept of coher-

(11) Definition. Kn = f | E{.

This is a true counterfactual based on the postulate (13), analogous

Since each of the two functions Cr 0 and Cred is definable on the

momentary inclinations—also with the second concept occurring in the

N o w we are ready to take the step from normative decision theory

Then we shall have in inductive logic, in analogy to the requirement

There is a class of axioms of inductive logic which I call axioms of

2 On this question, compare my [19536,pp. 193 f . ] reprinted in [1955,pp.21 f.].

I think that the theory of the J i - and ^-functions deserves the

(b) <g{H | Kn) = 0.99.

(a) is the statement of the evidence at hand, the same as in the

once Hume awakens our intellectual conscience, we recognize that

"Inductive logic" is here understood as the theory of probability

Dl-1. a. Two sets A, B are disjoint = D ( A and B have no common

Dl-2. Let X be a nonempty set and si be a class of subsets of X.

A field is closed under all finite sequences of Boolean operations (union,

Dl-3. Let X be a nonempty set, and si be a class of subsets of X.

We shall assume that for any language , the class of sentences in

For any language JSf to be used we shall proceed as follows. We

Dl-4. Let E and E' be propositions (events).

e. E is included in E' (or E strictly implies £") = D£ E <= E'.

B. THE BASIC AXIOMS

Al. Axiom of the lower bound: | E) ^ 0.

V(H C\ H'\E) = |Eftifl' |E r\H).

Tl-2. Some consequences of the axioms.

Proof. From A3: | £ ) = 1 - ^ { - H \ £).